Extended Data Fig. 3: Modulation bandwidth and frequency channel shifting with pseudorandom bit sequences.

a, Experimental set-ups. The input laser beam is modulated by either a sinusoidal signal from the port 1 of a vector network analyser (VNA) (bandwidth measurement) or an actual data stream that is generated by an arbitrary waveform generator (AWG) (eye diagram measurement). The input light is up-shifted and detected by a photodetector (PD) followed by either port 2 of the VNA (bandwidth measurement) or an oscilloscope (eye diagram measurement). The measurements are performed at a wavelength of 1,560 nm on the two-resonator device in Fig. 2b, c, in which the doublet splitting is 11.3 GHz due to optical dispersion (doublet splitting is 12.5 GHz at 1,601 nm). b, Measured modulation bandwidth of the device. The 3-dB modulation bandwidth is 2.2 GHz, corresponding to an optical bandwidth of 4.4 GHz. The modulation bandwidth is broadened to 4.1 GHz by increasing the microwave driving power from 126 mW to 398 mW. c, Measured eye diagrams when using actual data streams to modulate the input laser beam. The eye diagrams before shift are measured by setting the input laser beam off-resonance with the filter window centered to the input wavelength (filter bandwidth unchanged). By comparing the amplitude of the eye diagrams before and after swap, we found the loss is ~1 dB which is consistent with the on-chip loss (0.92 dB).